This Guardian article is interesting:
The research seems to be focused on popular western musical genres, although presumably it can be applied elsewhere. The idea is that you can mathematically model the popularity of music in the same way that the spread of a virus is modelled. I guess however you’d still need to make assumptions on, say, how any particular song will be received by the intended market. And this research was conducted quite a while ago, in the time when downloads were prevalent. But still very interesting, with possible implications for music industry predictions.